The NGA Director of Data and Digital Innovation predicts success with computer vision
Approximately one year ago, the National Geospatial-Intelligence Agency (NGA) began a seemingly herculean task: formulating an official point of view about artificial intelligence (AI), and summarizing that point of view—including NGA’s AI-related progress, activities, and objectives—into just a few short pages for easy sharing with employees, partners, and industry.
Titled “The Way of AI,” the finished document was completed approximately six months ago—just as OpenAI’s generative AI chatbot, ChatGPT, was coming to market, completely upending business as usual in enterprises across the public and private sectors alike.
Given the pace of change in AI and machine learning, that feels like millennia ago, according to NGA Director of Data and Digital Innovation Mark Munsell, who discussed NGA’s AI posture Wednesday afternoon during a keynote address at GEOINT 2023.
“It’s fascinating and awesome to see what’s already happened since [we] began writing this,” Munsell told an audience of GEOINT experts and enthusiasts at the Symposium’s Government Hub. “It’s really taken off.”
Of course, that’s no surprise to anyone who reads the news. In the last six months, generative AI has changed everything. So much so that some alarmed business leaders, policymakers, and technologists have called for a temporary moratorium on AI development.
Although IC leaders recognize and even share their concerns, there is no such moratorium at NGA, Munsell promised.
“We’re not pausing. The National Geospatial-Intelligence Agency is not pausing. And I don’t think the Department of Defense is pausing,” he said. “I say this because we’re still not good enough.”
According to Munsell, NGA’s AI interests lie mostly with computer vision—being able to detect and positively identify objects of interest with exquisite speed and accuracy.
“We’re not good enough,” he repeated. “That’s the state of where we’re at.”
That’s the bad news. The good news, Munsell said, is that the conditions are right for NGA to go from being good at AI to being better to ultimately being the best in the world.
“One of the things we’re most proud of is the environment that we’ve set up in our agency—in the community—to be able to bring in software faster,” explained Munsell, who said NGA now has the computing power and IT infrastructure it needs to achieve its most ambitious goals. “The setting is ripe. What we’ve been building toward is here.”
“I have combatant commanders that would expect the object identification to be 100% and the geolocation accuracy to be 2 meters or less so they can deliver a precision-guided munition. Those are the expectations,” said Munsell, who also discussed the transition of DoD’s Project Maven to NGA; his leadership agenda as head of NGA’s Data and Digital Innovation Directorate; and his excitement around large language models, which he thinks will eventually have implications for imagery as well as text. “Are we there? We’re not there. Will we get there? Absolutely. It’s inevitable…I don’t know when. I don’t know what the path is going to be. But we’re absolutely moving in that direction.”